“Find me a case that says…”

Anyone who started as a junior lawyer knows this request. It sounds simple, but in common law jurisdictions like Canada or the US, it usually leads down a rabbit hole. The senior lawyer asking always seemed certain the perfect case existed. Hours or days later, though, the answer was usually the same: there is no such case.

Unless you got lucky, the “right” case never existed, and your time went into building the best answer you could from the pieces you found: a footnote in a treatise, a first search with jurisdiction filters, and then, if you really needed to pin down results, Boolean strings like these:

• arbitrat! /s (unconscionab! or unenforceab!) /p employ! /s (compel! or motion or petition) /p (procedur! or substantiv!) & DA(aft 12-31-2025)
• ("restrictive covenant" OR non-compet! OR "non-solicitation") /p (employ! OR "former employee") AND (enforc! OR unenforceable OR "restraint of trade") AND (legitimate /5 interest!) AND NOT franchise

Then you read, weighed each case’s facts against the client’s situation, and searched again until you had enough pieces to make a sound argument. Finally, you wrote the memo (which could take several more hours on its own).

Before joining GC AI as an R&D Attorney, I was a litigator in Big Law. I know junior lawyers who avoided litigation placements because legal research “scared” them (their words, not mine). That's fair.

For me, though, this was the part of practicing law I enjoyed most. It felt like solving a puzzle, with room to be creative and build the strongest argument for the client while staying grounded in what the law says. It was also the part that took the longest.

In-house, you rarely have that much time for one question. So you might send it to outside counsel and pay for an associate's time researching and writing a memo, plus a partner's review. Or it might mean giving the business an answer on its timeline, before you've had the chance to confirm the state of the case law.

AI was supposed to fix this. But for many lawyers, it brought one word to mind: hallucinations. That concern is understandable. Case law research is hard, and models aim to please. So if you ask a model without the right data to “find a case that says X,” it may hand you a case that sounds too good to be true. That’s because it is: the case is made up.

We built GC AI’s US Case Law and International primary law research to solve that problem.

Starting with the data

You trust an answer when you know where it came from, so we started with the data. GC AI connects to dedicated legal databases. US Case Law draws on more than 14 million US court opinions, and the International database draws on case law and legislation from 30+ jurisdictions across Europe, Asia-Pacific, and Latin America. The data comes from original sources, such as court publishers and government websites, through Midpage AI and Moonlit, legal data providers. Both databases are updated daily, so what you reach through GC AI is current.

Every case GC AI cites is one it retrieved from those databases. The alternative is a model answering from memory, which is where the risk of made-up cases comes from. A model that has never read a decision can still produce a plausible case name, a citation in the right format, and a quote that sounds like a judge wrote it.

Searching by meaning

The search works differently too. A Boolean search matches your words, while GC AI searches by meaning. If you ask about non-competes after selling a business, it can find the case discussing “restrictive covenants ancillary to the transfer of goodwill,” without you having to guess how the court phrased it, or which keywords will “hit.”

Search by meaning is only as good as the data behind it, so we tested the data before relying on it. We ran evaluations on each source for completeness and coverage, and for how well its decisions are indexed, so GC AI can find the right case even when the court uses different words than you do.

Before relying on a case, GC AI reads the full decision, the way a careful associate pulls the opinion instead of trusting the headnote. GC AI’s case law data is also enriched with metadata that records how later courts have treated the case, so you can ask how later courts treated it.

Researching like a local

Connecting GC AI to the data was the first step. Getting it to research like a lawyer took a lot more. Translating those legal habits into the method GC AI follows was one of my favorite parts of building this launch.

When we built the International database, one of my first test questions was about a post-termination non-compete for an employee in Germany. An early answer went straight to court decisions, and they were real decisions on the right topic. But a German lawyer would have started with the Commercial Code, which generally requires the employer to pay compensation for the covenant to bind. The cases matter, but they only explain how courts apply the code.

I trained as a common law litigator, where the case itself is often the law. So we wrote down how a lawyer in each kind of system starts a research question, and GC AI now does the same. In civil law jurisdictions, it starts with the code. In common law jurisdictions, it starts with binding decisions from the highest court that governs your question. Either way, it explains how statute and case law fit together instead of assuming. R&D Attorneys then wrote research questions from real practice across the US, Europe, Asia-Pacific, and Latin America, each paired with what a good answer must include and what it must never claim. We graded GC AI’s outputs against these answer keys and kept refining its analysis until the answers read as if they came from someone who practices there.

The same searches, in plain language

Remember those two Boolean searches? Here is how you would ask GC AI the same questions today, the way you might ask a colleague down the hall:

• “Find decisions from 2026 onward where an employer moved to compel arbitration of an employment claim and the employee argued the arbitration agreement was unconscionable. Explain how courts treated procedural and substantive unconscionability.”
• “When will courts enforce a non-compete or non-solicit against a former employee? Focus on what counts as the employer's legitimate interest, and leave out franchise cases.”

Each prompt carries the same details the Boolean string did, but written in plain language. Behind each prompt, GC AI does the part that used to take the afternoon. It searches, reads the results, and searches again if those first results are thin, all with your jurisdiction and date range in mind.

In-house, you can go a step further and attach your own documents: “Here is our employment agreement. Is our non-solicit likely enforceable in California and in New York?” GC AI researches with your company context, relevant chat history in your conversation or Project, and the documents you attach, so the answer gets straight to what you need: what the law says, how it applies to your facts, and the practical takeaways for your company.

Your judgment in the loop

Getting to the answer now takes minutes. Deciding what to do with it is still your call, and that is where your judgment comes in. GC AI makes it easy to check its work. Every case GC AI cites in its response can be opened in full, where you can read the original text of the case right next to the answer, and verify it without switching tabs. For US cases, treatment flags show how later courts treated the decision. For international sources, you’re only one click away from viewing the case or code on its original source website. When GC AI considers several opinions in its analysis, those cases are set out in a summary table with the facts, issue, holding, and, most usefully, application: how each case applies to your context. Think of it as a headnote, but personalized.

The next time you find yourself tasked with a “find me a case that says…” question, you can respond quickly and with confidence. In-house, the business gets an answer on its timeline, backed by the law, and you’re the one who helped move it forward. And if the answer is still that there is no such case, at least now you’ll know by lunch.

P.S. Have a “find me a case” question of your own? Start a free trial or book a demo and try it now. Learn more about GC AI’s legal research.